Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

Journal of Reading Behavior

1980, Vol. XII, No. 1

TWO LETTER DISCRIMINATION SEQUENCES:


HIGH-CONFUSION-ALTERNATIVES FIRST VERSUS
LOW-CONFUSION-ALTERNATIVES FIRST1

Douglas W. Carnine
University of Oregon

Abstract. The procedure of sequencing visual discriminations (the letters b, d, p, and q)


in an easy to difficult progression was investigated. In a low-confusion-alternatives first
sequence only one choice in an inital match-to-sample task was similar to the target let-
ter. In a high-confusion-alternatives first sequence all the choices in the initial task were
similar to the target letter. Twenty preschoolers were randomly assigned to either a low-
confusion-alternatives first or a high-confusion-alternatives first group. The
preschoolers who began with the low-confusions alternatives reached criterion on a
subsequent successive match-to-sample task in significantly fewer trials, M = 31.0 ver-
sus 69.1.

Paired-associate learning is a complex process that consists of various subskills


(Keppel, 1968). If training procedures can be devised that facilitate learner per-
formance on a particular subskill, such as perceptual learning, the associational
learning process would be expedited. A particularly germane focus for perceptual train-
ing is letter discrimination (E. Gibson, J. Gibson, Pick & Osser, 1962; Staats, 1968). An
example of a training procedure was reported by Samuels (1973) who found that
children who learned to discriminate similar letters from each other (called distinctive-
feature training) learned letter-name associations in fewer trials than children who
received training on easier discriminations. Discrimination difficulty was defined ac-
cording to the structural similarity between a sample letter and the alternatives that ap-
peared in a match-to-sample task; e.g., b discriminated from d, p and q was a difficult
discrimination and was called a high-confusion alternatives treatment by Nelson and
Wein (1974) while ¿discriminated from c, o, and a was an easy discrimination to make
and was called a low-confusion alternatives treatment.
Nelson and Wein (1974) reported that although children in the low-confusion alter-
natives treatment reached criterion in fewer training trials, the children in the high-
confusion alternatives treatment made more correct responses in a posttest, which in-
cluded both high- and low-confusion alternatives. Similarly, Samuels (1973) reported
that although children who received distinctive-feature training required more trials to
42 Journal of Reading Behavior

reach criterion than children who did not receive the training, the distinctive-feature
training group learned to name four experimental letters in fewer trials. These results
supported Samuels' hypothesis that distinctive-feature training during the perceptual
learning phase of associational learning facilitated the S-R linkage.
However, a more careful inspection of Samuels' data (1973) raises questions about
the efficiency of the distinctive-feature training. When the trials from the distinctive-
feature training and from the letter-naming training are combined to form a total trials-
to-criterion measure, it appears that children required a comparable number of trials to
learn the letter names, whether or not they received distinctive-feature training. The
high-confusion alternatives group required an average of 24.37 trials to reach criterion
(12.27 on visual discrimination plus 12.10 on letter naming) while children in the low-
confusion alternatives group required an average of 21.69 trials (4.03 on visual
discrimination plus 17.66 on letter naming). A total trials-to-criterion measure is impor-
tant from a practical vantage point because the classroom teacher must consider the
total trials required to teach children letter names (or sounds), which includes trials re-
quired for visual-discrimination training and trials required for letter-name training.
Overall efficiency is not increased if savings inletter-name training are compensated for
by increased time requirements for distinctive-feature training.
The purpose of the present study was to investigate a sequencing procedure that
might be used to increase the efficiency of distinctive-feature training. Researchinvolv-
ing human (Marsh, 1967; Olson, 1963] and nonhuman subjects (Mackintosh, 1965;
Sutherland, Mackintosh, & Mackintosh, 1963; Schlosberg & Solomon, 1943) found that
acquisition of a single visual discrimination occurred in fewer trials when negative ex-
amples were sequenced to form an easy to difficult progression. Discrimination difficul-
ty was defined in terms of stimulus similarity along the dimension of color or shape. The
research did not investigate the procedure of sequencing a set of discriminations in an
easy to difficult progression with children as subjects. In the present study a set of
discriminations (the letters p, d, b, and q) was sequenced in an easy to difficult progres-
sion and children served as subjects. If sequencing a set of visual discriminations from
easy to difficult reduces the number of training trials required for children to reach
criterion in a perceptual learning task, children who first encountered match-to-sample
tasks with easy discriminations (low-confusion alternatives), and later difficult
discriminations (high-confusionaltemativesj.shouldreachcriterioninfewertrials than
children who first encountered the high-confusion alternatives. If sequencing
discriminations according to difficulty did speed up visual-discrimination training,
teachers could spend less time on distinctive-feature training, which would increase its
efficiency.
The possible efficiency of sequencing letter discriminations from easy to difficult
was suggested in two studies by Camine (1976) in which he sequenced letters in an easy-
to-difficult progressionbasedonresponse difficulty rather than onstructural similarity.
Letter-sound correspondences were introduced cumulatively (a letter appeared several
times in the context of previously introduced letters before a new letter was introduced)
according to a paired-associate paradigm. In the difficult-discrimination first group,
similar sounds were introduced one after another, while in the easy-discriminations
first group, dissimilar sounds were interpolated in the order of introduction to separate
similar sounds from each other. First graders in the easy-discriminationf irst group made
fewer training errors, and preschoolers in the easy-discrimination first group reached
Two Letter Discrimination Sequences 43

criterionin fewer training trials.


In the present study, discrimination difficulty was defined according to structural,
not acoustic similarity. The procedure was modeled after Samuels' (1973), in which
children were required to reach criterion on a simultaneous match-to-sample task, then
on a successive match-to-sample task, and finally on a letter-naming task. Since
students who responded to the successive match-to-sample task with high-confusion
alternatives reached criterion on the letter-naming task in significantly fewer trials, per-
formance on this successive match-to-sample task was selected as the criterion
measure. The appropriateness of a successive match-to-sample task is consistent with
Williams' (1969) finding that the most efficient method for focusing learner attention on
critical features was a delayed, match-to-sample task involving high-confusion alter-
natives.
Before children responded to the successive task in the present study, they had to
reach criterion on three different simultaneous match-to-sample tasks. In the high-
confusion alternatives first group, 75 % of the alternatives on the first match-to-sample
task were structurally similar; 25 % of the alternatives in the last task were similar. Con-
versely, in the low-confusion alternatives first group 25 % of the alternatives in the first
task were similar; 75 7o of the alternatives in the last task were similar. The hypothesis
was that children in the low-confusion first group would reach criterion on the suc-
cessive match-to-sample task in fewer trials than children in the high-confusion first
group.

METHOD

Subjects
From a population of 55 preschoolers between the ages of five and six, all children
who identified by sound or name no more than one of the experimental stimuli ( b, d, p, q,
a, o, and c) served as subj ects. The 32 children were randomly assigned to either the low-
confusion-alternatives first or the high-confusion-alternatives first group. Nine
children stopped attending the preschool, and three did not complete the experiment.
Ten subjects in each group—three boys and seven girls—completed the experiment. In
October, the mean age for the low-confusion alternatives first group was five years and
four months. The mean age for the high-confusion first group was five years and five
months.

Materials
Three training sets of simultaneous match-to-sample stimuli [A, B, and C) and the
criterion set (the successive set) were constructed by pasting xerox copies of two-inch
high letters on 8 Vi x 11 inch cardboard sheets (see Figure 1). In the three simultaneous
sets the sample letter appeared at the top of the sheet. Below the sample letter were the
correct choice (identical to the sample letter) and three alternatives. The four letters ap-
peared in random order below the sample letter. The sets were composed of highly
similar or less similar alternatives. The highly similar alternatives were b, d, p, q, The
less similar alternatives were a, c, o. While letters less similar to b, d, p, q, than o, a, c,
could have been used, these two sets of letters do differ substantially in terms of ap-
pearance. In relation to b, d, p, the letters a, c, and o have a mean similarity rating of 45,
44 Journal of Reading Behavior

P b d P q b q

xi o.
c a

XI CT
b c d a

CTO
0
CTO

CT03
p P
b 0 0 d P 0 c c c

d b d q d q P
a d b a c d d a d

CTO
CO CT

P 0

CTO
P 0 c d 0 0 q P 0 0 P

b p
a
CTO

P
CTO

b 0
Set A

p b d P q b q
XI
XI CT

a d d

XI 03
CTXl

CO CT

CO O.

CTXl
0 P 0 P
P d c b q d
O.

d b d q d q p
d b d b d

CO XI
CTCO

CTO.

O.CT
0 P
Û.CT

p
CTO

P q c d P d
c a

b p
a
CTÛ.

q
CTO

d b
Set B

p b d p q b q
XI CT

XI CT

CTCT

d d d
O.XI

Q.CT

CTÛ.
CTÛ.

d
CTXl

P
CTXl

d P q P P

d b d q d q p
O.

XI

b b
CTO.

b
CTO.
CTÛ.

O.CT

Û.CT

CTO.

P P q
q d d d
XI

d b
û.

b p
XI

d
CTÛ.
CTC

q
O-|

SetC

Figure 1. Match-to-sample stimuli for simultaneous sets A, B, and C.


Two Letter Discrimination Sequences 45

while b, d, p, have a mean similarity rating among themselves of 19 (Dunn-Rankin,


1968).
Simultaneous set A was considered a low-confusion alternative to the sample let-
ter. In simultaneous set B, two of the three alternatives were highly similar to the sample
letter. And in simultaneous set C, all three alternatives were highly similar to the sample
letter. The successive set was identical to simultaneous set Cexcept that the sample let-
ter in the successive set did not appear at the top of the card, but rather on the back of the
card.

Procedure
Preschoolers in both groups had to reach criterion (six consecutive correct
matches) on each set of stimuli before beginning the next set. The order for the low-
confusion first group was simultaneous set A, set B, and set Ciollowed by the criterion
(or successive) set. The order for the high-confusion first group was simultaneous set C,
set Band set A followed by the criterion set.
The experimenter presented the stimuli individually to each child in a quiet
testing room. Each session lasted no more than ten minutes. If a child had not
reached criterion on a particular stimulus set by the time the session ended, the child
began on that same set the next day. Children who did not reach criterion on a
simultaneous set within three days were dropped from the study. (Two children from
the low-confusion first group and one child from the high-confusion first group were
dropped.) After a child reached criterion on the third simultaneous set, the successive
set (or the criterion set) was presented. Children who did not reach criterion on the suc-
cessive set after 96 trials were assigned a score of 96 and training was terminated.
For the three simultaneous sets, the experimenter pointed to the sample letter and
said, "Look at this letter." After a two second pause, the experimenter said, "Now you
find the one that looks just like it. " If the child pointed to the sample le tter rather than one
of the four letters that appeared below the sample letter, the experimenter pointed to the
fourletters and said, "FindtheonedowflAerethatlooksjusth'keit."Thechi]dwasgiven
four seconds to respond. For the successive set the experimenter held up the card so that
the single letter was in view and said, "Look at this." After two seconds, the ex-
perimenter turned over the card, exposing the four alternatives, and said, "Now you
find the one that looks just like it." Again, the child was allowed four seconds to res-
pond.
The experimenter followed the same procedure for all four sets in reacting to child
responses: 1) If a child did not respond within four seconds, the experimenterprompted
the child by requesting the child to respond. The children were required to make a
response for each card presented. 2) Correct responses were praised. 3) After an incor-
rect response, the experimenterpointed to the correctletter.

RESULTS

The children in the low-confusion first group reached criterion in significantly


fewer trials on the successive task than children in the high-confusion first group. The
mean for the low-confusion first group was 31.0, SD = 37.7; the mean for the high-
confusion first group was 69.1, SD= 38.6, f(18) = 2.23,p< .025. In Table 1 child scores
46 Journal of Reading Behavior

within each treatment were grouped according to how many trials were required to
reach criterion on the successive task. The majority of children in the high-confusion
first group required the maximum number of trials to reach criterion; in fact, five of the
seven children in the 81-to-termination category never reached criterion—training was
terminated after the maximum number of trials. Two children from the low-confusion
first group did not reach criterion. The performance for the two groups on the three
simultaneous tasks was comparable. The mean for the low-confusion first group was
52.4, SI>= 27.6,andthemeanforthehigh-confusionfirstgroupwas51.8, SD= 31.7.

TABLE 1

Grouped Frequency Distribuction of


Trials on Successive
Match-to-Sample Task

Initial Discriminations
Trials Low Confusion High Confusion
0-20 6 2
21-40 1 1
41-60 1 1
61-80 0 0
81-termination 2 7

DISCUSSION

The results suggested that sequencing discriminations from easy to difficult [on the
basis of structural similarity) reduced the number of training trials r'equired in
distinctive-feature training. These results were consistent with those of Camine (1976),
who created an easy-to-difficult progression by separating similar sounds from each
otherintheirorderof introduction. The effectiveness of theeasytodifficultprogression
possibly results from the gradual development of stimulus differentiation. While initial
discriminations were relatively easy, they focused attention on relèvent dimensions
{presence of an appendage to a circular form). Althoughlater discriminations were more
difficult (focusing on the position of the appendage), students were prepared in that they
had already learned to attend to the relèvent dimension. Shifting learner attention to the
presence of the appendage and then to its position reduces a relatively difficult
discrimination for children to a two-step discrimination.
The implication for classroom applications is that distinctive-feature training might
proceed more rapidly if discriminations are introduced in an easy-to-difficult progres-
sion. This conclusion is tempered somewhat by the findings of Knafle and Legenza
(1978) that 1- and 2-inch stimuli are easier to learn than V* -inch stimuli. A replication of
the present study with normal size print couldresolve this question.
Two Letter Discrimination Sequences 47

The process of sequencing stimuli will often be complicated by the fact that some
members of the set, in this study letter-sound correspondences, may be structurally
and/or acoustically similar. The interaction of these two sources of similarity in deter-
mining discrimination difficulty and the sequencing implications are unclear and merit
further research. Additional research is also needed to determine the generalizability of
the sequencing procedure to other populations and other visual discrimination leaning
tasks. The research could incorporate the findings of Nelson and Peoples ( 1975) concer-
ning the importance of subject variables andresponse type.

FOOTNOTES
1
I would like to thank Nancy Prill for her assistance with the study. The study was supported in part
by grant OEG-0-70-4257 (286) from the United States Office of Education for the Training, Manage-
ment, and Research support for the Follow Through Districts (Engelmann-Becker Model).
Reprints may be obtained from Douglas W. Carnine, Department of Special Education, University
of Oregon, Eugene, Oregon, 97403.

REFERENCES

CARNINE, D.W. Similar sound separation and cumulative introduction in learning letter-sound
correspondences. Journal of Educational Research, 1976, 69, 368-372.
DUNN-RANKIN, P. The similarity of lower-case letters of the English alphabet. Journal of Verbal
Learning and Verbal Behavior, 1968, 1, 990-995.
GIBSON, E.J., GIBSON, J.J., PICK, A.D., & OSSER, H. A developmental study of the discrimina-
tion of letter-like forms. Journal of Comparative and Physiological Psychology, 1962, 55,
897-906.
KEPPEL, C. Verbal learning and memory. Annual Review of Psychology, 1968, 19, 169-202.
KNAFLE, J. D., & LEGENZA, A. External generalizability of inquiry involving artificial or-
thography. American Educational Research Journal, 1978, 15(2), 331-347.
MACKINTOSH, N. J. Selective attention in animal discrimination learning. Psychological
Bulletin, 1965, 64, 124-150.
MARSH, G. Intradimensional transfer of discrimination along the hue discrimination.
Psychonomic Science, 1967, 8, 411-412.
NELSON, R. O., & PEOPLES, A. A stimulus-response analysis of letter reversals, Journal of
Reading Behavior, 1975, 2, 329-340.
NELSON, R. O., & Wein, K. S. Training letter discrimination by presentation of high-confusion
veisuslow-confusion ahlternatives. Journal of EducationalPsychology, 1974, 66, 926-931.
OLSON, L. A, Concept attainment of high school sophomores. Journal of Educational Psychology,
1963, 54, 213-216.
SAMUELS, S. J. Effect of distinctive feature training on paired-associate learning. Journal of
Educational Psychology,1973, 64, 164-170.
SCHLOSBERG, H., & SOLOMON, R.L. Latency of response in a choice discrimination.Journalof
Experimental Psychology, 1943, 33, 22-39.
STAATS, A.W. Learning, language, andcognition. New York: Holt, Rinehart, and Winston, 1968.
SUTHERLAND, N. S., MACKINTOSH, H. N., &MACKINTOSH, J. Simultaneous discrimination
training of Octopus and transfer of discrimination along a continuum. Journal of Com-
parative and Physiological Psychology, 1963, 56, 150-156.
WILLIAMS, J. P. Training kindergarten children to discriminate letter-like forms. American
Educational Research Journal, 1969, 6, 501-514.

You might also like